The code in this repository accompanies the experiments performed in the paper Generalizable Adversarial Training via Spectral Normalization by Farnia*, Zhang*, and Tse (*equal contributors), which will be presented as a poster at ICLR 2019.
The repository contains the following:
- dl_spectral_normalization: Python deep learning module with spectral normalization code, code for building and training neural networks using TensorFlow, code for adversarially training networks, and example neural network architectures
- notebooks_figures: Contains scripts for generating all figures in the main text of the paper
- get_cifar10.py: Code for downloading and preprocessing datasets as described by Zhang et al. 2017
- train_network_template.ipynb: Example notebook for training a neural network using the dl_spectral_normalization module
The dl_spectral_normalization package can be installed via pip:
pip install dl_spectral_normalization
An example approach for accessing package contents is as follows:
# Imports utilities for building and training networks
from dl_spectral_normalization import dl_utils
# Import one of the provided neural network architectures: AlexNet
from dl_spectral_normalization.models import alexnet
# Import adversarial training methods
from dl_spectral_normalization import adversarial as adFor a more detailed tutorial, please refer to train_network_template.ipynb. For references on visualizing results, we provide several examples in notebooks_figures. We were able to run all of our experiments in an nvidia-docker image (tensorflow/tensorflow:latest-gpu running TensorFlow version 1.10.1).